CN100570619C - A Component Model Order Reduction Method for Product-Level Simulation - Google Patents

Abstract

A kind ofly belong to CAD area of computer aided Collaborative Product simulation technical field towards products grade simulated model of parts order reducing method, it is characterized in that, contain following steps: three-dimensional grid model and the compute mode of setting up parts with ANSYS software, extract the mass matrix and the stiffness matrix of model of parts, use Mathematica software again and in conjunction with the depression of order algorithm in the finite element analysis, according to resulting stiffness matrix, mass matrix, and the state during the structure maximum distortion under the typical load of setting, state during system balancing drops to the model of parts that m mode constitutes to the model of parts with n-dimensional vector.The present invention has improved the products grade simulated speed of complex product, has improved the degree of reusing of emulated data, and simple to operate, is easy to realize.

Description

A Component Model Order Reduction Method for Product-Level Simulation

technical field

The invention belongs to the technical field of computer-aided collaborative product simulation.

Background technique

Complex electromechanical products, such as machine tools, automobiles, robots, and aerospace vehicles, are usually a complex large system, including a large number of components or subsystems, involving multiple physical fields, and the performance of each part affects each other and is tightly coupled. . In the process of product design, design experts in many fields are needed to carry out design work according to the performance requirements of the product. In order to find problems in component design in the early stage of product design and reduce design changes, thereby reducing design costs and speeding up the design process, it is particularly important to introduce simulation technology in all aspects of the complex product design process.

However, due to the tight coupling between the various parts of many complex products, it is impossible to ensure that the product meets the design requirements simply by performing single-physics simulation on the performance of the components. An important means of design optimization. The so-called product-level simulation refers to the method of combining the models of parts or subsystems to form the overall model of the product, and using the overall model to simulate the product performance. However, due to the large number of parts and mutual constraints that make up the whole product, product-level simulation is difficult to implement.

An effective solution is to reduce the order of each component before product-level simulation, and then use the reduced-order component model to perform product-level simulation. Therefore, establishing a reasonable and effective component reduction model is the key to realizing product-level simulation. At present, the model reduction algorithm mainly focuses on the complex control system, and the more classic methods include: aggregation method, singular perturbation method, modal approximation method, Pade approximation method and so on. The mechanical research is mainly aimed at the simulation of the flow field model, especially the application of the POD reduction algorithm in the flow field, and there are few reduction algorithms for the component structure itself.

Contents of the invention

The purpose of the present invention is to provide a fast and accurate simplified calculation method for product-level simulation. On the basis of theoretical analysis, the model of the product parts is reduced based on the accurate simulation results, that is, the degree of freedom of the simulation model is reduced, thereby reducing the According to the requirements of computer software and hardware conditions, the calculation speed is accelerated, so as to ensure that the product-level simulation can be carried out quickly and accurately.

The present invention is characterized in that:

1. A component model reduction method for product-level simulation, characterized in that, the method is realized in the following steps in a computer:

Step (1). Building a component modeling module, a component model detailed simulation result extraction module, and a component model reduction module in the computer;

Step (2). Follow the steps below to create a 3D model of a component with the component modeling module:

Step (2.1) input the actual size and relevant parameters of the parts, use ANSYS software to build the three-dimensional model of the parts, and save;

Step (2.2) uses described ANSYS software to carry out tetrahedral grid division to the three-dimensional model that step 2.1 establishes, and preserves;

Step (2.3) uses the ANSYS software to add corresponding constraints to the three-dimensional mesh model of the parts set up in step 2.2 according to the actual situation of the parts operation, and save;

Step (3). Use the solution module Solution of the ANSYS software to simulate the part model built by the part modeling module. If the requirements are met, then perform the next step, otherwise return to step 2.1 and modify the parts parameters until the design requirements are met;

Step (4). Extract the simulation results obtained in step 3 according to the following steps successively;

Step (4.1) uses the modal analysis (Modal) in the solution module Solution, selects the block Lanczos (BlockLanczos) method to calculate the modal, and saves;

Step (4.2) extracts the stiffness matrix and the mass matrix of the component model with the ANSYS software according to the modality of the component model obtained in step 4.1;

Step (5). Read in described mass matrix and stiffness matrix, call Mathematica software, carry out order reduction operation to parts and components according to the following algorithm:

其中n维向量x_eqm为已知系统平衡状态，为一组设定的n维线性无关向量基(即模态向量)，

表示将向量x_off与向量

点乘后得出的数值绝对值，

为向量基

的2-范数，将a_i从大到小排列，选用前m个a_i对应的m个模态编程构建零部件降阶模型，则零部件自由度从n降低至m。The state of the three-dimensional mesh model of the parts at the time of the maximum deformation of the structure under the set typical load conditions is represented by an n-dimensional vector x _max , where n is the number of degrees of freedom; definition: x _off = x _max -x _eqm ,

Among them, the n-dimensional vector x _eqm is the known equilibrium state of the system, For a set of n-dimensional linearly independent vector bases (ie, modal vectors),

Indicates to combine the vector x _off with the vector

The absolute value of the value obtained after dot multiplication,

is a vector basis

The 2-norm of , arrange the a _i from large to small, and select the m modal programming corresponding to the first m a _i to construct the component reduction model, then the degree of freedom of the component is reduced from n to m.

The advantages of this component model reduction method for product-level simulation are:

(1) Using JSP-based network means to realize the model reduction algorithm, which provides necessary preparations for the realization of multi-person collaborative sharing of simulation results, the realization of standardized management of simulation processes, data, and results, and improves the product-level simulation speed of complex products. Effectively shorten the product development cycle.

(2) Improve the reuse of simulation data. The reduced-order model is built on the basis of accurate simulation results of parts, which improves the reusability of accurate simulation results of parts and reduces the cost of reduced-order modeling.

(3) The operation is simple and easy to implement. Mature commercial software is used for accurate simulation of component models, and the construction of reduced-order models can be realized by simple programming using high-level languages.

Description of drawings

Fig. 1. Structural frame diagram of the present invention;

Figure 2. The relationship between the time-consuming solution and the degrees of freedom after order reduction;

Figure 3. The mesh division diagram of the wing model;

Figure 4. Comparison of the reduced-order solution and the non-reduced-order solution of the wing model;

Figure 5. Error bar plot of the reduced-order solution of the wing model.

Detailed ways

The simulation method is a model reduction method for component structures, which includes the following three modules, as shown in Figure 1, which are:

1. Component Modeling Module

This module is mainly to establish the 3D mesh model of the parts, and provide the basic model for other modules to complete detailed simulation, order reduction and other calculations. The steps involved are as follows:

Step 1.1 establishes the three-dimensional solid model of the component. According to the actual size and material of the parts, through input devices such as mouse and keyboard, use ANSYS software to manually build its model, and the established 3D model is saved in model.sat format.

Step 1.2 meshes the component model. In ANSYS software, for the 3D solid modeling of parts established in step 1.1, the 3D model is divided into tetrahedral meshes by mouse, keyboard and other input devices in the order from line to surface and from surface to body, so as to achieve continuous system discrete For the purpose of optimization, the grid model of the parts is obtained and saved as a model.db format file.

Step 1.3 adds constraints to the component model. In the ANSYS software, add corresponding constraints to the 3D mesh model of the parts established in step 1.2 according to the actual operation of the parts, and save it as a model.db format file. The constraints are exported by the ANSYS software and saved as Forces.txt.

2. Part model detailed simulation result extraction module

This module mainly uses the solution module of ANSYS software to carry out detailed simulation of the component model built by the component modeling module. If the design requirements are met, follow the steps below to extract the detailed simulation results of the component model, so that the component model If it is not satisfied, return to the component modeling module, and modify the component parameters until the detailed simulation meets the design requirements:

Step 2.1 calculates the modes of the component model. Mode is an inherent vibration characteristic of a vibrating system, generally including frequency, mode shape, damping, etc. Since parts are divided into multiple small units during finite element analysis, there are multiple degrees of freedom, so multiple vibration modes (also called mode vectors) can appear, and multiple natural frequencies (also called mode vectors) can appear at the same time. Modal frequency), the mode of the structure is only related to the parameters of the structure itself, and has nothing to do with external force and damping. In the ANSYS software, apply the "Modal" analysis in the "Solution" function module to the component model established in step 1, and select the "Block Lanczos" method to calculate the modal, and the modal analysis results of ANSYS are stored in model.full in the file.

Step 2.2 extracts the stiffness matrix of the component model. Select the "List/Files/Binary Files" command in the ANSYS software, select "Matrix" in the pop-up dialog box, import the modēl.full file obtained in step 2.1, select "Stiffness" in the "Matrix to write" box, click OK, and extract zero The stiffness matrix of the component model is a Stiff.txt file.

Step 2.3 extracts the mass matrix of the component model. Select the "List/Files/Binary Files" command in the ANSYS software, select "Matrix" in the pop-up dialog box, import the model.full file obtained in step 2.1, select "Mass" in the "Matrix to write" box, click OK, and extract zero The mass matrix of the component model is a Mass.txt file.

3. Component model reduction module

This module calls the relevant functions of Mathematica software through programming to perform order reduction calculations on parts and components, which is the core module of this method.

Step 3.1 Determine the algorithm for model reduction. The principle of order reduction is as follows: When analyzing solid structures, the common method is finite element analysis. Assuming that a structure has n degrees of freedom after meshing, the state of the system can be represented by an n-dimensional vector x. Usually, in order to truly simulate the performance of the structure, the divided mesh will be relatively fine, and the corresponding n will be very large. In order to perform a more accurate simulation with fewer degrees of freedom, the following approximate transformation can be done:

为一组已知的m维线性无关向量基，q_i为相应迭加系数，如此系统自由度便可从n降阶到m，当m＜＜n时，仿真过程便可大大加快。由以上分析可知，模型降阶过程中一个非常关键的问题就是如何寻找适当的向量基 Among them, the n-dimensional vector x _eqm is the equilibrium state of the system,

is a set of known m-dimensional linear independent vector bases, and q _i is the corresponding superposition coefficient, so that the degree of freedom of the system can be reduced from n to m. When m<<n, the simulation process can be greatly accelerated. From the above analysis, it can be seen that a very critical issue in the process of model reduction is how to find an appropriate vector basis

在保证计算精度的前提下，利用这些选取的模态来进一步构建零部件的降阶模型，从而降低模型的自由度。Since the elastic body is divided into multiple small units during the finite element analysis, various mode shapes (mode vectors) appear, and there are multiple natural frequencies (mode frequencies). Accurate model, adopt the following modal impact evaluation method based on maximum deformation to select the modal that has the greatest impact on structural performance

Under the premise of ensuring the calculation accuracy, these selected modes are used to further construct the reduced-order model of the parts, thereby reducing the degree of freedom of the model.

Assuming that under a typical load condition, the state of the maximum deformation of the structure is x _max ,

Definition: x _off = x _max - x _eqm ,

为一组设定的n维线性无关向量基(即模态向量)，

表示将向量x_off与向量

点乘后得出的数值取绝对值，

为向量基的2-范数，将a_i从大到小排列，则前m个a_i对应的模态对结构的变形影响最大，选择这m个模态构建零部件降阶模型，模型自由度则从n降至m。Among them, the n-dimensional vector x _eqm is the known equilibrium state of the system,

For a set of n-dimensional linearly independent vector bases (ie, modal vectors),

Indicates to combine the vector x _off with the vector

The value obtained after dot multiplication takes the absolute value,

is a vector basis The 2-norm of the a _i is arranged from large to small, then the modes corresponding to the first m a _i have the greatest influence on the deformation of the structure, and these m modes are selected to build a component reduction model, and the model degrees of freedom are from n drops to m.

Step 3.2 establishes the business logic processing class of the component model reduction module. Use the Java language to write the business logic processing class, and realize the order reduction algorithm in step 3.1 by calling the relevant functions in the Mathematica software. It mainly includes writeModel, read in the modal analysis results; writeMassMatrixGif, read in the mass matrix, and do a simple analysis; writeStiffMatrixGif, read in the stiffness matrix, and do a simple analysis; eigenSystem, solve the eigenvalue and eigenvector; nSolve2, use the order reduction The method numerically solves the zero initial condition problem; nSolveWithoutReduction3 numerically solves the zero initial condition problem with the non-reduction method; error compares the error between the reduction and non-reduction cases.

Step 3.3 Create the page display of the component model reduction module. Use the Java language to write the page display of JSP, encapsulate it in the package "ROM", and import the business logic processing class of the component model reduction module completed in step 3.2 on the web page and the components from the component modeling module, component by writing a program The load, mass matrix, and stiffness matrix input of the model detailed simulation result extraction module realize the order reduction function of the component model. Among them, InputMatrix.jsp is the program entry, which is responsible for uploading the detailed simulation result data to the server; ShowGraph.jsp is for visually displaying the uploaded data, and allowing the user to select some options, which will affect the algorithm processing of the reduction analysis; Solution.jsp displays the corresponding solution results according to the solution options above. The solution results that can be displayed include: reduced-order analytical solution (the analytical solution should not be calculated when the problem is complex), non-reduced analytical solution (the analytical solution should not be calculated when the problem is complex), and the reduced-order value under zero initial conditions solution, non-reduced numerical solution under zero initial conditions, and compare the error caused by reduced order; CloseMath.jsp is to close the Mathematica engine running in the background of the server to save server resources and provide more simulation information for the next user of CPU resources.

Step 3.4 uses Eclipse software to package the business logic processing class and page display established in steps 3.2 and 3.3 into a package in war format, and name it ROM.war.

In order to better understand the technical solution of the present invention, the above algorithm is applied to the analysis of wing models of common components in the aerospace field for further description.

Step 1 establishes a 3D solid model of the wing. According to the actual size and material of the parts, through input devices such as mouse and keyboard, use ANSYS software to manually build its model, and the established 3D model is saved in wing.sat format.

Step 2 Mesh the wing model. In the ANSYS software, for the 3D solid modeling of the wing established in step 1, divide the 3D model into tetrahedral meshes through input devices such as mouse and keyboard according to the order from line to surface and from surface to body, and divide it into With 460 movable nodes, the grid model of the parts is obtained, as shown in Figure 3, which is saved as a wing.db format file.

Step 3 Add constraints to the wing model. In ANSYS software, add corresponding constraints to the wing three-dimensional mesh model established in step 2 according to the actual operation of the parts, and save it as wing.db format file. The constraints are exported by ANSYS software and saved as Forces.txt.

Step 4 Compute the modality of the wing model. In the ANSYS software, apply the "Modal" analysis in the "Solution" function module to the part model established in step 3, and select the "Block Lanczos" method to calculate the mode, and the results of the modal analysis by ANSYS are stored in wing.full in the file.

Step 5 extracts the stiffness matrix of the wing model. Select the "List/Files/Binary Files" command in the ANSYS software, select "Matrix" in the pop-up dialog box, import the wing.full file obtained in step 4, select "Stiffness" in the "Matrix to write" box, click OK, and extract the machine The stiffness matrix of the wing model is Stiff.txt file.

Step 6 extracts the mass matrix of the wing model. Select the "List/Files/Binary Files" command in the ANSYS software, select "Matrix" in the pop-up dialog box, import the wing.full file obtained in step 4, select "Mass" in the "Matrix to write" box, click OK, and extract the machine The stiffness matrix of the wing model is a Mass.txt file.

Step 7: Installation of component model reduction module. The server computer starts the J2SDK software, imports the ROM.war file into the webapps folder under the Apache Tomcat software installation path, and restarts Apache Tomcat to directly apply the component model reduction module established above. Just enter the server ip: port number/ROM in the address bar of the client computer to use it.

Step 8: Input of component model reduction module. Input the Forces.txt file obtained in step 3, the wing.full file obtained in step 4, the Stiff.txt file obtained in step 5, and the Mass.txt file obtained in step 6 into the corresponding dialog box of the component model reduction module, and according to You need to select the corresponding calculation options, such as zero initial condition reduced-order numerical solution, zero initial condition non-reduced numerical solution, error analysis, etc., click OK to submit the input.

Step 9: Output of component model reduction module. According to the input in step 8, the part model reduction module will output the corresponding calculation results, and finally the CloseMath page will pop up, and the user can choose to close the Mathematica engine running in the background of the server to save server resources and provide more information for the next user's simulation. of CPU resources.

Figure 2 shows the time-consuming calculation of the reduced order of the wing model from 1 to 100. The curve is approximately a parabola, indicating that the difficulty of solving the problem is roughly in a quadratic relationship with the reduced order, and As the order increases, the time-consuming becomes longer; Fig. 4 is an image of a solution changing with time in the numerical solutions (that is, the displacement of each point) obtained by the two models, and the two are in good agreement; Fig. 5 is the The corresponding error bar chart after the wing model is reduced to order 40, it can be seen that the relative errors are all less than 3%. ; In addition, in terms of time comparison, the calculation time without de-escalation is 97.41 seconds, and the calculation time after de-escalation is 4.86 seconds.

Claims (1)

1. A component model reduction method for product-level simulation, characterized in that, the method is realized in the following steps in a computer: Step (1). Building a component modeling module, a component model detailed simulation result extraction module, and a component model reduction module in the computer; Step (2). Follow the steps below to create a 3D model of a component with the component modeling module: Step (2.1) input the actual size and relevant parameters of the parts, use ANSYS software to build the three-dimensional model of the parts, and save; Step (2.2) uses described ANSYS software to divide the three-dimensional model that step (2.1) establishes into a tetrahedron grid, and save; Step (2.3) uses said ANSYS software to add corresponding constraints according to the actual situation of parts operation to the parts three-dimensional mesh model that step (2.2) establishes, and save; Step (3). Use the solution module Solution of the ANSYS software to simulate the component model built by the component modeling module, if it meets the requirements, then perform the next step, otherwise return to step (2.1), modify Component parameters until the design requirements are met; Step (4). Follow the steps below to extract the simulation results obtained in step (3): Step (4.1) uses the modal analysis (Modal) in the solution module Solution, selects the block Lanczos (BlockLanczos) method to calculate the modal, and saves; Step (4.2) according to the mode of the part model that step (4.1) obtains, extract the stiffness matrix and the mass matrix of part model with described ANSYS software; Step (5). Read in described mass matrix and stiffness matrix, call Mathematica software, carry out order reduction operation to parts and components according to the following algorithm: The state of the three-dimensional mesh model of the parts at the time of the maximum deformation of the structure under the set typical load conditions is represented by an n-dimensional vector x _max , where n is the number of degrees of freedom; definition: x _off = x _max -x _eqm ,

Among them, the n-dimensional vector x _eqm is the known equilibrium state of the system, For a set of n-dimensional linearly independent vector bases, that is, modal vectors,

Indicates to combine the vector x _off with the vector The absolute value of the value obtained after dot multiplication,

is a vector basis

The 2-norm of , arrange the a _i from large to small, and select the m modal programming corresponding to the first m a _i to construct the component reduction model, then the degree of freedom of the component is reduced from n to m.

Images